A kinome siRNA screen identifies HGS as a potential target for liver cancers with oncogenic mutations in CTNNB1

Aberrant activation of the Wnt/β-catenin pathway is a major and frequent event in liver cancer, but inhibition of oncogenic β-catenin signaling has proven challenging. The identification of genes that are synthetically lethal in β-catenin-activated cancer cells would provide new targets for therapeutic drug design.

R E S E A R C H A R T I C L E Open Access

A kinome siRNA screen identifies HGS as a

potential target for liver cancers with

Frédéric Canal1,2,3,4,5*, Elodie Anthony6, Aurianne Lescure6, Elaine Del Nery6, Jacques Camonis6, Franck Perez6, Bruno Ragazzon1,2,3and Christine Perret1,2,3,4

Abstract

Background: Aberrant activation of the Wnt/β-catenin pathway is a major and frequent event in liver cancer, but inhibition of oncogenicβ-catenin signaling has proven challenging The identification of genes that are

synthetically lethal inβ-catenin-activated cancer cells would provide new targets for therapeutic drug design Methods: We transfected the parental HuH6 hepatoblastoma cell line with a doxycycline-inducible shRNA against CTNNB1 (gene coding for β-catenin) to obtain an isogenic cell line pair with or without aberrant β-catenin signaling Using this hepatoblastoma isogenic cell line pair, we performed a human kinome-wide siRNA screen to identify

synthetic lethal interactions with oncogenicCTNNB1 The phenotypic readouts of the screen were cell proliferation, cell cycle arrest and apoptosis, which were assessed by image-based analysis In addition, apoptosis was assessed by flow cytometric experiments and immunoblotting The potential synthetic lethal relationship between candidates genes identified in the screen and oncogenicCTNNB1 was also investigated in a different cellular context, a colorectal HCT116 isogenic cell line pair

Results: We first determined the experimental conditions that led to the efficient expression of shRNA against CTNNB1 and maximal reduction of β-catenin signaling activity in response to doxycycline treatment Following high throughput screening in which 687 genes coding for kinases and proteins related to kinases (such as

pseudokinases and phosphatases) were targeted, we identified 52 genes required for HuH6 survival The silencing

of five of these genes selectively impaired the viability of HuH6 cells with highβ-catenin signaling: HGS, STRADA, FES, BRAF and PKMYT1 Among these candidates, HGS depletion had the strongest inhibitory effect on cell growth and led to apoptosis specifically in HuH6 with highβ-catenin activity, while HuH6 with low β-catenin activity were spared In addition,HGS was identified as a potential synthetic lethal partner of oncogenic CTNNB1 in the HCT116 colorectal isogenic cell line pair

Conclusions: These results demonstrate the existence of crosstalk betweenβ-catenin signaling and HGS

Importantly, HGS depletion specifically affected cells with uncontrolledβ-catenin signaling activity in two

different types of cancer (Hepatoblastoma HuH6 and colorectal HCT116), and thus may represent a new potential target for novel therapeutic strategies in liver and colorectal cancer

Keywords:β-catenin, Synthetic lethality, High throughput screening, Liver cancer, HGS

* Correspondence: frederic.canal94@gmail.com

1 Department Development, Reproduction and Cancer, INSERM U1016,

Institut Cochin, 24, rue du Faubourg Saint-Jacques, 75014 Paris, France

2 CNRS UMR8104, Paris, France

Full list of author information is available at the end of the article

© 2015 Canal et al Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made The Creative Commons Public Domain Dedication waiver

Genetic alterations and modifications of the tumor

en-vironment often lead to the apparition of weaknesses

specific to the tumor that could be exploited

therapeut-ically [1] For instance, the discovery of synthetic lethal

(SL) interactions in cancer cells offers a framework for

the design of highly selective drugs SL interactions

occur between two genes when mutation of one gene

alone does not alter cellular fitness, but mutation of

both causes cell death [2] Therefore, the discovery of SL

interactions involving an “undruggable” oncogene may

lead to the identification of new potential therapeutic

tar-gets for cancer treatment This has been well illustrated by

the LS interaction between BRCA and PARP1 Indeed,

PARP1 inhibitors show promising activity in clinical trials

of breast, ovarian and other cancers associated withBRCA

mutations [3]

Deregulation of the Wnt/β-catenin pathway, which is a

key developmental biology signaling pathway, is a major

event in liver cancer and colorectal tumorigenesis [4, 5],

which were the 2ndand 4thleading causes of death by

can-cer worldwide in 2012, respectively (WHO) Indeed, more

that 50 % of hepatoblastoma (HB) and a third of

hepatocel-lular carcinoma (HCC) display aberrant activation of Wnt/

catenin signaling caused by stabilizing mutations of

β-catenin in the CTNNB1 gene [4, 6], while mutations in

APC, which lead to the ectopic activation of Wnt/β-catenin

signaling, are considered the major initiating event in

colo-rectal cancer (CRC) [5, 7] Thus, the Wnt/β-catenin

path-way has become a prime target for cancer research

However, despite intensive research during the past decade,

the production of molecules effective against cancers

asso-ciated with uncontrolled Wnt/β-catenin signaling activity

has proven challenging [8] Hence, the identification of SL

partners of oncogenicβ-catenin is a promising strategy for

the discovery of new therapeutic targets for liver cancer

Here, we used an siRNA kinome library to perform a

high throughput (HT) SL screen of HuH6 isogenic HB

cell lines, and we examined the effect of knockdown on

cell proliferation, the frequency of mitotic events and

in-duction of apoptosis The depletion of transcripts of five

genes by siRNA led to lethality only in a cellular context

characterized by the aberrant activation of the Wnt/β −

catenin signaling pathway due to a CTNNB1 activating

mutation One of these genes (HGS) is an SL partner of

oncogenic β-catenin in colorectal HCT116 cancer cells,

suggesting that the LS interaction between HGS and

CTNNB1 is not limited to liver cancer

Methods

Cell culture, transfection and generation of stable shRNA

clones

Human hepatoblastoma HuH6 cells were grown in

Dulbelcco’s modified Eagle’s medium (DMEM, Gibco,

Life Technologies, Carlsbad, CA) with 10 % fetal bovine serum and 100 U/ml penicillin/streptomycin Colorectal carcinoma HCT116 cells were cultivated in McCoy’s medium, with 10 % fetal bovine serum, at 37 °C in 5 %

CO2 Parental HuH6 cells were transfected with pTER-β-catenin plasmid using Lipofectamine 2000 (Life Technolo-gies) to generate HuH6shCTNNB1 cells [9] Positive clones were selected following the culture of cells in 5μg/ml puro-mycin for 4 weeks Isolated colonies were picked using cloning rings and clones were amplified for 6 weeks and stored in liquid nitrogen prior to further analysis

Reporter assay

The TOPflash/FOPflash reporter plasmids (Millipore, Billerica, MA) were used to determine β-catenin-induced TCF/LEF transcriptional activity TOPflash is a reporter plasmid containing two sets of three copies of wild-type TCF binding sites driven by the thymidine kinase minimal promoter located upstream from a luciferase reporter gene FOPflash contains mutated TCF binding sites and is used as a negative control for TOPflash activity HUH6 and HUH6shCTNNB1were culti-vated in the presence or absence of 2μg/ml of doxycyc-line for 72 h and transfected with reporter plasmids using Lipofectamine2000 in triplicate in accordance with the manufacturer’s instructions The pRL-TK plasmid (Promega, Madison, WI) was co-transfected to control for transfection efficiency Forty-eight hours after trans-fection, Luciferase activity was measured with the Dual-Luciferase reporter assay system (Promega)

Real Time quantitative PCR

Total RNA was isolated with TriZol reagent according

to the manufacturer’s instructions (Life Technologies) Reverse transcription was performed from 1 μg of total RNA with the Transcriptor First Strand cDNA Synthesis Kit (Roche Diagnostics, Basel, Switzerland) and random hexamer primers PCR amplification was performed on the LightCycler 480 system with SYBRGreen PCR mix (Roche Diagnostic) and the following primers: HGS for-ward 5’- CTCCTGTTGGAGACAGATTGGG -3’ and H

GS reverse 5’- GTGTGGGTTCTTGTCGTTGAC -3’, 18S forward 5’-GTAACCCGTTGAACCCCATT-3’ and 18S reverse 5’-CCATCCAATCGGTAGTAGCG-3’, CTNNB1 forward 5’- GCTTTCAGTTGAGCTGACCA-3’ and CTN NB1 reverse 5’-GCTTTCAGTTGAGCTGACCA-3’ or Axin2 forward 5’- TGTCTTAAAGGTCTTGAGGGTTG AC-3’ and Axin 2 reverse 5’- CAACAGATCATCCCAT CCAACA-3’

Transcriptome analysis

After validating RNA quality with the Bioanalyzer 2100 (using Agilent RNA6000 nano chip kit), 50 ng of total RNA was reverse transcribed with the Ovation PicoSL

WTA System V2 (NuGEN, San Carlos, CA) Briefly, the

resulting double-stranded cDNA was used for

amplifica-tion based on SPIA technology After purificaamplifica-tion

according to the manufacturer’s protocol, 2.5 μg of

single-stranded DNA was fragmented and labeled with

biotin using the Encore Biotin Module (NuGEN)

Frag-ment size was verified with the Bioanalyzer 2100, cDNA

was then hybridized to GeneChip® human Gene 1.0 ST

(Affymetrix) at 45 °C for 17 h After overnight

hybridization, the chips were washed on the fluidic

sta-tion FS450 according to the manufacturer’s protocol

(Affymetrix, Santa Clara, CA) and scanned with the

GCS3000 7G The image was then analyzed with

Expres-sion Console software (Affymetrix) to obtain raw data

(cel files) and metrics for quality control The evaluation

of some of these metrics and the distribution of raw data

showed no experimental outliers RMA normalization

was performed with R

High-content siRNA screening

A kinome siRNA library of 2748 siRNAs targeting 687

genes (four siRNAs per gene) was obtained from Qiagen

We also included a negative siRNA control (anti-

lucifer-ase, GL2) and a positive lethal siRNA (anti-KIF11) on

each plate A total of 750 cells were seeded on a

384-well plate (View Plates, Perkin Elmer) in 40 μL of

DMEM medium supplemented with 10 % fetal bovine

serum and 1 % penicillin/streptomycin Cells were

trans-fected the day after with 10 nM siRNA using

INTER-FERin, (0.1 μl per well; Polyplus Transfection, Illkirch,

France) Transfection efficiency was estimated by

com-paring 4,6-diamidino-2-phenylindole (DAPI) staining

be-tween cells transfected with negative control siGL2 and

those transfected with siKIF11 Cells were fixed 72 h

after transfection with 4 % (w/v) formaldehyde for

15 min and washed with phosphate buffered saline

(PBS) Cells were next quenched with 0.05 M NH4Cl

and permeabilized first with a PBS solution containing

0.2 % BSA and 0.05 % saponin, and subsequently with

0.5 % Triton-X100 Cells were then incubated for 60 min

with rabbit primary antibody anti-cleaved Caspase 3

(1:500; Sigma-Aldrich) and mouse anti-Ki67 (1:500;

Millipore) Next, cells were washed twice in PBS and

in-cubated with Alexa Fluor 488- or Cy3 coupled secondary

antibodies (Jackson ImmunoResearch) Nuclei were

stained with 0.2μg/ml DAPI Images were acquired with

an INCell2000 automated wide-field system (GE

Health-care,) at 10x magnification (Nikon 10X/0.45, Plan Apo,

CFI/60) Images were analyzed with the INCell Analyzer

workstation software (GE Healthcare) and the mean

fluorescence intensity of nuclear Ki67 (a proliferation

marker) and cleaved-caspase3 antigens was quantified in

each cell Three independent experiments were

per-formed for each treatment

Data analysis and hit calling

Positive hits for each gene were identified as follows Data were first transformed with log or logit functions B-score normalization was then applied to each replicate, separ-ately, and included corrections for plates, rows and col-umns [10, 11] Median and median absolute deviation (MAD) were computed and used to compute Robust Z-scores (RZ-Z-scores) for each sample, according to the for-mula: score = (value - median)/(1.4826 * Median MAD) [12] RZ-scores were calculated for the comparison of each siRNA against the GL2 negative control population

A gene was identified as a‘hit’, if the RZ-score for at least two of four siRNAs was > 2 or < -2 in at least two of three replicates Selected hits were further reordered and vali-dated with four siRNAs per gene (Qiagen), for HUH6 and HUH6shCTNNB1cell lines treated or not treated with 2μg/

ml doxycycline

Crystal violet assay

HuH6shCTNNB1cells were cultured with 2μg/ml of doxy-cycline for 48 h prior to siRNA transfection and were kept in the presence of doxycycline for the following

72 h HCTT116, HuH6 and HuH6shCTNNB1 cells were plated at 0.1 × 106 cells per well on 6-well plates and transfected with siRNA as indicated Seventy-two hours after transfection, cells were washed once in PBS and in-cubated in a crystal violet solution (0.5 % crystal violet,

20 % methanol) for 10 min at room temperature Plates were washed once with PBS and twice by immersion in water Cells were then incubated in 1 % SDS solution until the dye was completely solubilized and absorbance was read at 570 nm

Apoptosis assay

Cells were harvested and centrifuged at 200 g for 5 min and the pellet was resuspended in 500 μl of culture medium containing 200nM 3,3'-Dihexyloxacarbocyanine iodide (DiOC6(3)) and 7.5 μM propidium iodide (PI), (Life Technologies) Cells were incubated for 20 min at

37 °C and samples were analysed using a BD LSR For-tessa flow cytometer (BD Biosciences, San Jose, CA)

Cell cycle assay

Cells were harvested and fixed with ice cold ethanol

70 % for 20 min, stained with 7.5 μM PI solution con-taining 100 μg/ml of RNAseI, and then analysis were performed with a BD LSR Fortessa flow cytometer

Immunoblot analysis

Cells were washed in ice-cold PBS and lysed in RIPA buffer (Sigma-Aldrich) containing 1X of complete prote-ase inhibitor cocktail (Roche Diagnostics) and centri-fuged at 13,000 g for 10 min at 4 °C Total proteins were resolved by SDS-PAGE, transferred to nitrocellulose and

blocked with 5 % BSA Blots were separately incubated

overnight at 4 °C with specific primary antibodies, including

1:1000 anti-HGS/HRS (from Bethyl Laboratories,

Mont-gomery, TX), 1:5000 anti-β-catenin (BD Transduction

Laboratories, Franklin Lakes, NJ), 1:5000 anti-γ tubulin and

1:5000 anti-β-Actin (Sigma-Aldrich) After further washing

and incubating with corresponding secondary antibodies,

the blots were developed by enhanced chemiluminescence

(Thermo Fisher Scientific, Waltham, MA)

Results and discussion

Identification of liver cancer cell lines with highβ-catenin

signaling activity

Oncogenic mutations of β-catenin are a frequent event

in liver cancer In this study, we used an siRNA-based

loss-of-function screen to identify kinases and proteins

related to kinases that are specifically required for the

survival of cells in which β-catenin signaling is

deregu-lated First, we measured β-catenin-induced Tcf/Lef

transcriptional activity in liver cancer cell lines with a

dual-luciferase assay to identify pertinent cell lines for

HT-screening We selected liver cancer cell lines with

CTNNB1S37C

and CTNNB1G34V

found in SNU398 and HuH6 cells, respectively, or a CTNNB1ΔW25_I140

dele-tion found in HepG2 cells), and a cell line with a WT

CTNNB1 gene (HCC HuH7 cells) HeLa and Human

Embryonic Kidney 293 T cells (HEK293T) were used as

negative controls Only background activity was detected

in HeLa and HEK293T cells, whereas HuH7 cells showed

modest but significant β-catenin co-transcriptional

activ-ity, despite the WT status ofβ-catenin (Fig 1), which may

be caused by autocrine Wnt signaling in HuH7 cells [13]

Cell lines harboringCTNNB1 mutations showed great

dis-parity in basal β-catenin signaling activity SNU398 cells

showed rather weak β-catenin transactivation (3.6 times

higher than background levels, similar to HuH7 cells),

whereas activity was 8- and 36-fold higher than

back-ground levels in HepG2 and HuH6 cells, respectively

(Fig 1) Therefore, HuH6 cells were subsequently selected

for HT-screening as a reference cell line with high

β-catenin signaling activity

Generation of a stable HuH6 cell line with

doxocycline-inducible sh-RNA againstCTNNB1

We sought to identify SL interactions involving

onco-genic β-catenin activity; therefore, we transfected HuH6

cells with pTER-sh-β-catenin encoding a Doxycycline

(Dox)-regulated shRNA against CTNNB1 [9], to obtain

HuH6 cells with low β-catenin activity HuH6 clones

containing a stably integrated shRNA vector were

selected by culture in the presence of 5 μg/ml

puro-mycin for four weeks We then analyzed the abundance

of β-catenin expression in selected clones treated with

Dox by western blotting The HuH6_5G clone showed the lowest levels of β-catenin (Fig 2a) The abundance

lower in cells treated for 5 days with Dox than in control cells, confirming the efficiency of the shRNA (Fig 2b) Additional experiments showed that β-catenin signaling was strongly impaired in the Dox-treated HuH6_5G clone RT-qPCR experiments showed thatAXIN2 (a dir-ect target of TCF/LEF/β-catenin complex [14, 15]) mRNA levels were 80 % lower in HuH6_5G cells treated for 5 days with Dox than in untreated (no Dox) control cells (Fig 2c) In addition, we compared gene expression profiles between untreated and Dox-treated HuH6_5G cells by Affymetrix DNA microarrays: several genes that are known to be activated byβ-catenin were significantly down-regulated in cells expressing shRNA against CTNNB1 (Table 1 and Additional file 1: Table S1), indi-cating that β-catenin signaling activity was impaired in these cells Finally, TCF/LEF luciferase reporter activity was substantially lower in Dox-treated HuH6_5G than

in control cells (Fig 2d), confirming that β-catenin sig-naling is weak when shRNA expression is induced in re-sponse to Dox treatment in HuH6_5G cells

Surprisingly, the almost complete inhibition of β-catenin signaling activity did not appear to correlate with β-catenin protein levels (which were only 40 % lower in Dox-treated cells than in control cells) This may be explained by the dual role of β-catenin at the plasma membrane, where it has a structural function in cadherin‐based adherens junctions [16] Indeed, the remainingβ-catenin protein in Dox-treated cells is likely

to be trapped in cadherin-based adherent junctions and

Fig 1 β-Catenin transactivation level in human HCC and HB cell lines Hepatocellular carcinoma cell lines (HepG2 and SNU398), hepatoblastoma cell lines (HuH6 and HuH7), Human embryonic kidney 293 T cell line and human cervical carcinoma HeLa cell line were transfected with either TOPFlash luciferase (black rows) or FOPFlash luciferase (gray rows) reporter plasmids Firefly luciferase activity was assessed 24 h after transfection in whole cell extracts and normalized to Renillia luciferase activity

thus not able to activate genes transcription (β-catenin signaling is close to basal level in Dox-treated cells) Fur-thermore, we observed thatβ-catenin signaling was simi-lar between parental HuH6 and untreated-HuH6_5G cells (compare Figs 1 and 2d) For clarity, the HuH6_5G clone

is hereafter named HuH6shCTNNB1 Thus, parental Huh6 and HuH6shCTNNB1(without Dox) were used as reference cells with a high β-catenin signaling activity and Dox-treated HuH6shCTNNB1 cells were used as reference cells with lowβ-catenin signaling activity

The impairment of β-catenin signaling in Dox-treated HuH6shCTNNB1 cells was associated with a substantial cell growth defect when cells were seeded at very low density (Additional file 2: Figure S1-A), whereas cell growth was not significantly affected when cells were seeded at medium density (Additional file 2: Figure S1-B) In order to explain the slow increase of Dox-treated HuH6shCTNNB1in culture,

we performed cell cycle analysis by measuring DNA con-tent in HuH6shCTNNB1 treated or not with Dox by flow cytometry after PI staining (Additional file 2: Figure S1 C-D) We did not detect any significant difference between Dox-treated and untreated HuH6shCTNNB1 when seeded at medium density (Additional file 2: Figure S1-D) However,

at low cell density we observed an increase in G0/G1 cells

untreated cells, with a concomitant decrease in G2/M cells (Additional file 2: Figure S1-C) These data suggest that

Fig 2 Down regulation of β-catenin expression and inhibition of β-catenin signaling in HuH6 pTER-sh-β-catenin transfectant cells a HuH6 cells were transfected with pTER-sh- β-catenin and selected with 2 μg/ml puromycin for 4 weeks After amplification of isolated colonies, resultant HuH6pTER-sh-β-catenin transfectants were cultivated in the presence or absence of 2 μg/ml doxycycline for 72 h and the abundance of HuH6pTER-sh-β-catenin in whole cell extracts was assessed by immunoblotting b Clone 5G was cultivated in the presence or absence of 2 μg/ml doxycycline for the indicated times and β-catenin mRNA levels were analyzed by RT-qPCR c- Clone 5G was cultivated in the presence or absence of 2 μg/ml of doxycycline for 72 h and Axin2 mRNA levels were analyzed by RT-qPCR d Clone 5G was cultivated in the presence or absence of 2 μg/ml doxycycline for

72 h and β-catenin/TCF transcriptional activity was assessed by the luciferase reporter assay

Table 1 Target genes activated by Wnt/β-catenin signaling and

down-regulated in HuH6shCTNNB1in response to Dox treatment

Gene expression was analyzed with Affymetrix DNA microarrays

and the expression profile of HuH6shCTNNB1cells treated with Dox

for 2 days was compared with that of untreated HuH6shCTNNB1cells

Note thatCTNNB1 showed a -0.9 Δlog2 in response to shRNA

in-duction with Dox

G0/G1 phase of cell cycle is longer in Dox-treated

HuH6shCTNNB1when seeded at low density Yet, this

mod-est difference in cell cycle cannot fully explain the drastic

decrease in Dox-treated HuH6shCTNNB1growth rate shown

in Additional file 2: Figure S1-A Therefore, we evaluated

apoptosis by DiOC6(3) and PI co-staining DiOC6(3) is a

membrane-permeable lipophilic cationic fluorochrome that

is used as a probe for mitochondrial transmembrane

poten-tial (Δψm) In living cells, DiOC6(3) accumulates in

mito-chondria and low DiOC6(3) staining reflects a collapse of

Δψm Dissipation of mitochondrial transmembrane

poten-tial is a sensitive marker of early apoptotic events In

addition, PI stains only cells whose plasma membrane

in-tegrity is compromised Thus, cells that showed low

stain-ing for both DiOC6(3)(loss ofΔψm) and PI (loss of plasma

membrane integrity) were considered as apoptotic cells

Interestingly, we observed a significant increased in number

of apoptotic cells (8.7 % Vs 17.3 %) only in Dox-treated

HuH6shCTNNB1when seeded at low density (Additional file

2: Figure S1-E) Thus, prolongated G0/G1 phase and

sig-nificant increase of apoptosis could explain the slow growth

rate of Dox-treated HuH6shCTNNB1 when seeded at low

density Interestingly, these results are consistent with those

of Chan et al who showed that the inactivation of mutated

CTNNB1 in the colon cancer HCT116 cell line did not

modify the growth of cells passaged under routine

condi-tions, although it significantly decreased colony-forming

ability when plated at low density [17]

HT siRNA screen of HuH6 cells

Kinases play a crucial role in the regulation of

Wnt/β-catenin pathway (and in some cases can be targeted by

chemicals inhibitors) [18]; therefore, we carried out an

siRNA kinome-wide loss-of-function screen (Qiagen,

687 genes - Additional file 3: Table S2) in the HuH6

par-ental cell line, to identify kinases required for cell

viabil-ity in cells with high β-catenin (Fig 3a) Each gene was

targeted by four independent siRNAs, resulting in a

li-brary of 2748 siRNAs Positive (KIF11 siRNA) and

nega-tive (GL2 siRNA) controls were also added We

examined the effect of silencing on cell growth by

high-content immunofluorescence imaging with DAPI

stain-ing and specific antibodies against Ki67 (cell

prolifera-tion marker) and cleaved caspase 3 (apoptosis)

Each screen was repeated three times to obtain

bio-logical replicates Results are expressed as median

Z-scores for each siRNA All siRNAs with a medianZ-score

>2 or <−2 were considered significant hits, and genes with

at least two (out of four) ‘hit’ siRNAs were selected as

candidate genes (Fig 3a) Overall, 52 kinases important

for cell proliferation and/or cell viability were identified

(Additional file 3: Table S2), which represents 7.7 % of

identified kinases and associated proteins tested during

the screening procedure These results are consistent with

previous reports involving other cancer cell lines [19] As expected, survival kinases such as those involved in cell cycle progression (e.g AURKB, CDK3, CDK4, CDK5, CDK7, PLK3, PKMYT1) or in the control of cell growth (e.g ADK, RPS6KA2) were identified as hits in our pri-mary screen However, depletion of such cell cycle or cell growth rate related-kinases is likely to be deleterious in wild type cells as well

Identification of SL interactions involving oncogenic β-catenin signaling in HuH6 cells

We then carried out an SL counter-screen with siRNA against the 52 survival kinases to identify kinases whose inhibition would affect the viability of HuH6 cells with oncogenic β-catenin activity, but not that of HuH6 cells with low β-catenin activity (sh-in-duced knock-down of β-catenin) A scheme of the counter-screen is illustrated in Fig 3b Briefly, HuH6 and HuH6shCTNNB1 were cultivated for 5 days in the presence or absence of Dox, and screened with siRNA targeting the 52 kinases identified in the primary screen Thus, this counter-screen included three ex-perimental conditions with high β-catenin activity (HuH6, HuH6 + Dox, and HuH6shCTNNB1 cells) and one with low β-catenin activity (HuH6sh CTNNB1+ Dox cells) Proliferation index was slightly lower in HuH6shCTNNB1+ Dox cells than in the other three conditions and was strongly influenced by cell density (Additional file 2: Figure S1); therefore, the number

of cells/well during the seeding step was optimized to obtain the same number of cells for each condition during the transfection step, with similar growth rate

in each experimental condition (Additional file 4: Figure S2) Cell transfection, cell staining, immuno-fluorescence imaging and statistical analysis were car-ried out with similar methods used in the primary screen and three independent experiments were per-formed Genes were classified as a potential SL part-ner of the activated form of β-catenin if (i)|median robust Z-score| > 2 for at least two phenotypes in

HuH6 + Dox, HuH6shCTNNB1), and for at least two siRNAs; and (ii) knockdown had no significant effect (|median robust Z-score| < 2) in cells with low β-ca-tenin signaling (HuH6shCTNNB1+ Dox) for all phenotypes analyzed This counter-screen identified five genes that are potential SL partners of oncogenicβ-catenin (Table 2): HGS (hepatocyte growth factor-regulated tyrosine kinase substrate), STRADA (STE20-related kinase adaptor alpha), FES (feline sarcoma oncogene), BRAF (v-raf murine sarcoma viral oncogene homolog B) and PKMYT1 (protein kinase, membrane associated tyro-sine/threonine 1)

HGS is required for the survival of liver cancer cell lines

with oncogenic mutations inCTNNB1

We investigated biological effect of the five identified

candidates First, we performed the knockdown of

HGS, STRADA, FES, BRAF and PKMYT1 by RNAi

experiments in HuH6shCTNNB1 (high β-catenin

signal-ing) or in Dox-treated HuH6shCTNNB1 (low β-catenin

signaling) and assessed cell survival by crystal violet

staining (Additional file 5: Figure S3) We observed

that HGS depletion had the stongest inhibitory effect on

cell growth specifically in cells with highβ-catenin activity

However, the crystal violet assay is not a sensitive method for the estimation of cell survival Indeed, we cannot dis-criminate between adherent living and adherent dead cells with this assay By flow cytometric experiments, we mea-sured a two-fold increase of apoptotic cells following HGS knockdown in highβ-catenin signaling HuH6shCTNNB1

com-pared to scramble siRNA, while siRNA againstHGS did not modify proportion of apoptotic cells in HuH6shCTNNB1+ Dox (Fig 4b and Additional file 6: Figure S4) These data were confirmed by Western blot analysis of cleaved-caspase 3 expression, a specific marker of apoptotic cells

a

b

Fig 3 siRNA screen based on immunofluorescence imaging a Identification of kinases required for HuH6 viability A kinome siRNA library was used to transfect HuH6 cells: 687 genes were individually targeted (four siRNAs per gene) and three independent experiments were performed Cells were immunolabeled 72 h after transfection with anti-cleaved caspase 3 (apoptosis marker) and anti-Ki67 (cell proliferation markers) antibodies and nuclei were labeled with DAPI (cell count) before the acquisition of images and readout of phenotypes Three phenotypes were analyzed: (i) cell numbers through DAPI staining, (ii) Apoptosis and (iii) G0-phase arrest with Ki67 All siRNAs with median Z-scores >2 or < −2 were considered significant hits, and genes with at least two (out of four) ‘hit’ siRNAs were selected as candidate genes Following this screen,

52 genes were identified as necessary for HuH6 survival or proliferation The 52 outliers are listed in Table S2, additional Excel file b Counter-screen: Investigation of synthetic lethality relationship between candidate genes and stabilized β-catenin HuH6 and HuH6 shCTNNB1 were cultured in the presence or absence of 2 μg/ml doxycycline 72 h, and transfected with a siRNA library including the 52 outliers identified in the primary screen (four siRNAs per gene) Each screen was repeated three times to obtain biological replicates Transfection was carried out and phenotypes were assessed as described in A Outliers were called if |robust Z-score| > 2 for at least two phenotypes and two out of four siRNAs, only in cells with high β-catenin signaling Following the counter-screen, five genes (listed in Table 2) were identified as having a potential lethal synthetic relationship with

mutant β-catenin

Indeed, cleaved-caspase 3 was only detected after HGS

knockdown performed in high β-catenin signaling cells

(Fig 4c) Finally, to show that deleterious effect of HGS

depletion truly depends on oncogenic β-catenin mutant,

we performedHGS knockdown in HuH7 which is a

hepa-toblastoma cell line with wild-typeβ-catenin HGS protein

expression was efficiently impaired by HGS_1 siRNA in

HuH7 (Fig 4e) However, HGS depletion did not induce

apoptosis as indicated by the proportion of DiOC6(3)/PI

double negative cells measured by flow cytometry (Fig 4d)

and by the absence of cleaved-caspase 3 expression

(Fig 4e) Altogether, our data showed thatHGS is SL

part-ner of oncogenicβ-catenin and its depletion induces

apo-tosis in hepatoblastoma cell lines

HGS is required for the survival of CRC cell lines with

oncogenic mutations inCTNNB1

We decided to analyze SL interactions with oncogenic

CTNNB1 in a different cellular context, the human

colorectal cancer cell line HCT116, to determine the

specificity of the five potential SL partners of

activated-β-catenin Parental HCT116 is heterozygous for CTNNB1

(CTNNB1WT/del45

) and has one WT allele and one mutant allele with a 3-bp deletion that eliminates the serine

resi-due at codon 45, leading to synthesis of mutantβ-catenin

which is constitutively active [17] In parallel, we used

genetically modified HCT116_CTNNB1

WT/-cells with low β-catenin activity in which the CTNNB1del45

allele is invalidated [17] We first performed siRNA experiments

for the five targets identified during the SL screen to verify

the efficiency of silencing in HCT116 cells by RT-qPCR

(Additional file 7: Figures S5) We then estimated cell

sur-vival by the crystal violet assay 72 h after transfection of

HCT116WT/del45and HCT116WT/-cells with the most

ef-ficient siRNA against one of the five potential SL partners

of oncogenicβ-catenin Depletion of STRADA, FES, BRAF

and PKMYT1 affected cell survival in both HCT116WT/

del45

and HCT116WT/-, whereas HGS knockdown

signifi-cantly impaired cell survival only in HCT116WT/del45(high

β-catenin signaling cells) (Additional file 8: Figure S6) Indeed,HGS knockdown efficiently impaired the accumu-lation of HGS mRNA and protein in both cell lines (Fig 5a-b); however, the density of HCT116WT/del45 cells depleted ofHGS was 42 % lower than that of control cells, whereas HCT116WT/-cell density was unaffected byHGS depletion (Fig 5c) Moreover, HGS knockdown in HCT116WT/del45 leads to a significant increase in apop-totic cells number (11.6 % Vs 16.6 %), suggesting that HGS expression is required for HCT116WT/del45

survival (Fig 5d) Intriguingly, the abundance ofHGS mRNA was higher in HCT116WT/del45cells than in HCT116WT/-cells (Fig 5a), which suggest that HGS expression is in part positively regulated byβ-catenin in HCT116 cells (directly

or indirectly) However, HGS protein levels were not sig-nificantly different between HCT116WT/del45 and HCT116WT/-cells (Fig 5b) Thus, the potential regulation

ofHGS by β-catenin cannot fully explain the SL relation-ship between these two genes Our results obtained in the colorectal HCT116 cell line support the existence of an

SL relationship betweenHGS and oncogenic mutations in β-catenin and identify HGS as a potential therapeutic tar-get for cancers with mutations inCTNNB1, which is a fre-quent event in liver cancer

HGS (also called HRS for Hepatocyte growth factor-regulated tyrosine kinase substrate) is an essential protein that belongs to multi-vesicular bodies (MVB)/Lysosome complex, which is required to sort membrane proteins and direct them toward either degradation in lysosomes

or recycling at the plasma membrane [20, 21] Interest-ingly, Toyoshima et al showed that depletion of HGS in fibroblasts led to a defect in E-cadherin degradation and accumulation of E-cadherin at the plasma membrane, which in turn trapped β-catenin at the membrane, pre-venting its translocation to the nucleus In addition,HGS depletion attenuates the cell growth, tumorigenesis and metastatic potential of Hela cells [22] Therefore, it is plausible that HGS depletion in HuH6 cells leads to the accumulation of E-cadherin at the plasma membrane,

Table 2 List of the five genes identified by HT screening as important for the cellular fitness of highβ-catenin signaling cells (HuH6) but not for lowβ-catenin signaling (HuH6sh CTNNB1+ Dox) cells All siRNAs with a median RZ-scores <−2 or >2 were considered hits Genes were considered as hits if at least two of the four siRNAs were siRNA hits, for at least two phenotypes (cell proliferation, apoptosis or mitotic events) in highβ-catenin signaling cells (HuH6, HuH6 + Dox, HuH6sh CTNNB1), but not in lowβ-catenin signaling cells (HuH6sh CTNNB1+ Dox).

The RZ-score for the best siRNA hits/gene is shown

Genes

Hits

HuH6 HuH6+

Dox

HuH6shCTNNB1 HuH6shCTNNB

+ Dox

HuH6 HuH6+

Dox

HuH6shCTNNB1 HuH6shCTNNB

+ dox

HuH6 HuH6+

Dox

HuH6shCTNNB1 HuH6shCTNNB

+ Dox

b

c

Fig 4 HGS knockdown impairs cell viability specifically in hepatoblastoma cells with high β-catenin signaling a HuH6 shCTNNB1 cells were cultured

in the presence or absence of 2 μg/ml doxycycline for 96 h, seeded in 12-well plate and transfected with scramble or siRNA targeting specifically HGS and cultured for 72 h Untransfected cells (siRNAimax lipofectamine alone) were also included as a control Cell density was estimated 72 h after cell transfection by crystal violet staining b Cells were cultured and transfected as in A and were stained with DiOC6 (3) and PI Proportion of apoptotic cells was determined by flow cytometric experiment Note that apoptotic cells showed low labelling for both DiOC6 (3) and PI Means of three independent experiments are shown c Cells were treated as in A and cleaved caspase3, HGS and β − actin protein levels were examined by immunoblotting d HuH7 were transfected with scramble or siRNA targeting specifically HGS and cultured for 72 h Untransfected cells (siRNAimax lipofectamine alone) were also included as a control Proportion of apoptotic cells was determined by flow cytometric experiment after DiOC6 (3) /IP staining e HuH7 cells were transfected as in D and cleaved caspase3, HGS and β − actin protein levels were examined by immunoblotting Protein extract from HuH7 treated with 2 μg/ml doxorubicin for 16 h was added as positive control for cleaved caspase3 expression

which may trap the stabilized form of β-catenin,

thus impairing signaling through this pathway and

attenuating its oncogenic effect However, HGS is

also a key regulator of the molecular machinery

in-volved in receptor sorting and recycling; therefore,

HGS depletion may affect several signaling pathways

Hence, the exact nature of the SL relationship

be-tween oncogenic CTNNB1 and HGS requires further

investigation

Conclusions

In conclusion, we performed a siRNA screen that led to the identification of HGS as a potential SL partner of oncogenic CTNNB1 To ensure high confidence in the hits selected with our approach, we measured three fitness variables (cell proliferation index, induction of apoptosis and number of mitotic events) [23] in two dif-ferent isogenic cell line pairs from difdif-ferent cellular con-texts (liver cancer and CRC) Thus, our findings strongly

a

b

c

d

Fig 5 HGS knockdown impairs cell viability specifically in HCT116 cells with oncogenic mutations in β-catenin HCT116 wt/del

( CTNNB1 S45del

heterozygous) and HCT116 WT/- ( CTNNB1 WT only) cells were transfected with scramble or siRNA targeting specifically HGS and cultivated for 48 h Untransfected cells were also included as a control a HGS mRNA levels were assessed by RT-qPCR and normalized to the abundance of 18S RNA The mean ± SD of triplicate samples is presented b The abundance of HGS and γ-tubulin protein was examined by immunobloting c Cell proliferation was visualized by crystal violet staining 72 h after HGS knockdown Dye was solubilized in 1 % SDS solution and absorbance was read at 570 nm and quantified The mean ±

SD of triplicate samples is presented d Assessment of apoptosis with DiOC6 (3) and IP staining The mean ± SD of three independent experiments

is presented